lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <173665959558.1629214.16724136597211810729.stgit@devnote2>
Date: Sun, 12 Jan 2025 14:26:35 +0900
From: "Masami Hiramatsu (Google)" <mhiramat@...nel.org>
To: Steven Rostedt <rostedt@...dmis.org>,
	Jiri Olsa <olsajiri@...il.com>
Cc: Alexei Starovoitov <alexei.starovoitov@...il.com>,
	Florent Revest <revest@...omium.org>,
	linux-trace-kernel@...r.kernel.org,
	LKML <linux-kernel@...r.kernel.org>,
	Martin KaFai Lau <martin.lau@...ux.dev>,
	bpf <bpf@...r.kernel.org>,
	Alexei Starovoitov <ast@...nel.org>,
	Alan Maguire <alan.maguire@...cle.com>,
	Mark Rutland <mark.rutland@....com>,
	Masami Hiramatsu <mhiramat@...nel.org>
Subject: [PATCH] fgraph: Move trace_clock_local() for return time to function_graph tracer

From: Masami Hiramatsu (Google) <mhiramat@...nel.org>

Since the ftrace_graph_ret::rettime is only referred in the function_graph
tracer, the trace_clock_local() call in fgraph is just an overhead for
other fgraph users.

Move the trace_clock_local() for recording return time to function_graph
tracer and the rettime field is just zeroed in the fgraph side.
That rettime field is updated by one of the function_graph tracer and
cached for other function_graph tracer instances.

According to Jiri's report[1], removing this function will gain fprobe
performance ~27%.

[1] https://lore.kernel.org/all/Z3aSuql3fnXMVMoM@krava/

Reported-by: Jiri Olsa <olsajiri@...il.com>
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@...nel.org>
---
 include/linux/ftrace.h               |   11 +++++++++++
 kernel/trace/fgraph.c                |    2 +-
 kernel/trace/trace_functions_graph.c |    2 ++
 3 files changed, 14 insertions(+), 1 deletion(-)

diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
index 07092dfb21a4..a2fb7360d6e2 100644
--- a/include/linux/ftrace.h
+++ b/include/linux/ftrace.h
@@ -1155,6 +1155,17 @@ struct ftrace_graph_ret {
 	unsigned long long rettime;
 } __packed;
 
+static inline void ftrace_graph_ret_record_time(struct ftrace_graph_ret *trace)
+{
+	if (!trace->rettime)
+		trace->rettime = trace_clock_local();
+}
+
+static inline void ftrace_graph_ret_init_time(struct ftrace_graph_ret *trace)
+{
+	trace->rettime = 0;
+}
+
 struct fgraph_ops;
 
 /* Type of the callback handlers for tracing function graph*/
diff --git a/kernel/trace/fgraph.c b/kernel/trace/fgraph.c
index c928527251e3..bc1e5f0493b6 100644
--- a/kernel/trace/fgraph.c
+++ b/kernel/trace/fgraph.c
@@ -826,7 +826,7 @@ __ftrace_return_to_handler(struct ftrace_regs *fregs, unsigned long frame_pointe
 		return (unsigned long)panic;
 	}
 
-	trace.rettime = trace_clock_local();
+	ftrace_graph_ret_init_time(&trace);
 	if (fregs)
 		ftrace_regs_set_instruction_pointer(fregs, ret);
 
diff --git a/kernel/trace/trace_functions_graph.c b/kernel/trace/trace_functions_graph.c
index d0e4f412c298..7e4d91d42d8e 100644
--- a/kernel/trace/trace_functions_graph.c
+++ b/kernel/trace/trace_functions_graph.c
@@ -327,6 +327,7 @@ void trace_graph_return(struct ftrace_graph_ret *trace,
 	int size;
 	int cpu;
 
+	ftrace_graph_ret_record_time(trace);
 	ftrace_graph_addr_finish(gops, trace);
 
 	if (*task_var & TRACE_GRAPH_NOTRACE) {
@@ -361,6 +362,7 @@ static void trace_graph_thresh_return(struct ftrace_graph_ret *trace,
 	struct fgraph_times *ftimes;
 	int size;
 
+	ftrace_graph_ret_record_time(trace);
 	ftrace_graph_addr_finish(gops, trace);
 
 	if (trace_recursion_test(TRACE_GRAPH_NOTRACE_BIT)) {


Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ